✦
Summary
Gemini 3.8 Flash is Google reasoning model for software engineering, tool use, and extended agent workflows. The release emphasizes stronger multi-step problem solving and professional task execution at the introductory token pricing of its predecessor, giving developers a model aimed at both capability and throughput.
Key Features
The release focuses on long-horizon coding, agentic work, and multi-step reasoning.
Iterative tool calls support checking and completing complex tasks.
Lower effort settings can reduce reasoning-token overhead.
The model may use additional tokens to improve difficult-task results.
The introductory API input rate is $0.75 per million tokens.
The introductory API output rate is $3.75 per million tokens.
It is a specialized variant for vulnerability detection and automated patching.
The Fairwind Program provides access to selected trusted defenders.
About Gemini 3.8 Flash
The model improves coding and reasoning through further training and more persistent tool-driven work. On difficult requests it may spend additional tokens checking and refining results, while lower effort settings provide a way to control overhead. The related Flash Cyber variant is tailored to security analysis.
Gemini 3.8 Flash is useful for applications involving autonomous development, research, and structured professional work. Hosted access and usage conditions vary by Google product. Flash Cyber is separately restricted to trusted defenders through the Fairwind Program and should not be assumed available through ordinary Flash access.